Re: [Opensg-users] OpenSG2: compiling with gcc-4.6 patch

[Michael B. <mba...@cr...>]

On 04/14/2011 12:01 AM, Gerrit Voß wrote:
>
> Hi,
>
> On Tue, 2011-04-12 at 15:19 -0500, Carsten Neumann wrote:
>> 	Hello Michael,
>>
>> On 04/11/2011 12:03 PM, Michael Babcock wrote:
>
>>> The OpenSG headers, at least, also almost compile with g++-4.6
>>> -std=c++0x, except for one error:
>>>
>>> /mnt/hdpsbin/dev/OpenSG-2.0/include/OpenSG/OSGFieldContainer.inl: In
>>> function 'ContainerPtr OSG::convertToCurrentAspect(ContainerPtr) [with
>>> ContainerPtr = OSG::MemoryObject*]':
>>> /mnt/hdpsbin/dev/OpenSG-2.0/include/OpenSG/OSGMemObjPointerMFieldCommon.inl:504:66:
>>> instantiated from here
>>> /mnt/hdpsbin/dev/OpenSG-2.0/include/OpenSG/OSGFieldContainer.inl:528:58:
>>> error: 'class OSG::MemoryObject' has no member named 'getAspectPtr'
>>>
>>> Where it seems to be passing a MemoryObject* where a FieldContainer* is
>>> expected.
>>
>> this one, should also not be specific to gcc 4.6 or c++0x mode, but I'm
>> not sure what the correct fix is. MemoryObjects don't have the multi
>> aspect copies FieldContainers have, so there is no meaningful way to get
>> the aspect local copy of one.
>> I'm guessing the MemObj fields would share the MemoryObject across
>> aspects. Gerrit, what's the intended semantics here?
>
> I'll look into it.
>
> kind regards
>    gerrit

FYI, a newer version of g++-4.6 (Debian 4.6.0-4, SVN 20110419 r172584) 
no longer complains about this in -std=c++0x mode. So perhaps it was a 
bug in gcc that was fixed.

I'm not sure why it's not an issue, perhaps that code is not actually 
instantiated, or is instantiated in a non-obvious manner such that the 
compiler knows the MemoryObject is always a FieldContainer. Sorry about 
the false alarm.